CONTROL Resonant (PPSA34547)

[hugodearaujo]

Existing thread

• I searched this category for the title ID and no thread exists for this title on any platform.

Title

CONTROL Resonant (PPSA34547)

Content version tested

01.300.300

AnyPS5 commit

main 6005185, alone and with #7300.

System tested

Arch Linux (kernel 7.2.9), AMD Radeon RX 9070 XT with RADV (Mesa 26.2.4), built from source with GCC 16.2.1, Release

Furthest point

Menu

What is shown

The photosensitivity warning, the Remedy end-user license agreement and privacy policy (they scroll with the right stick and accept with Cross), the display calibration screen with its three rendered reference scenes, and First Time Setup, at about 28 FPS. Cross, the D-pad and the right stick work. Audio was not checked (SDL dummy driver).

Current blocker

None found yet: the runs so far ended on First Time Setup, 7 minutes in, with nothing in the log; what follows it is not tested yet.

Needed to get here on main 6005185:

• fix(ampr): return EBUSY when a command does not fit and EPERM before a buffer is bound #7300 (merged as 582de4cc1, so current main needs nothing else to get here): without it, sceAmprAprCommandBufferReadFile returns 0x8002001C once a 0x400-byte APR command buffer is full, the title only resubmits on 0x80020010, 116 of its 141 effects are never read, and it dereferences a null effect (shapeengine.rfx) 7 s after start (SIGSEGV, nothing in the log).

How it was run

libcohtml.Prospero.prx, libRenoirCore.PS5.prx, MediaDecoders.Prospero.prx and nakama-sdk.prx are in the root of app0, so the relinker was run with --module-dir . (and --registry). No environment variables beyond the input file (APS5_PAD_INPUT_FILE).

AI-assisted: yes (Claude Code)

[ps54all]

What comes after First Time Setup, from Windows (RTX 3080 10 GB, NVIDIA 610.74, WinLibs GCC 15.2, Release; same content version 01.300.300, relinked with --module-dir .). Our build is a local test branch (main 82f0c9a plus open PRs), not clean main, so the stops below are the ones left on that branch.

Driving: First Time Setup, then the title screen and the main menu (~15 FPS here). Enter on the first entry starts New Game.

Stops after New Game, in order:

1. VGT_SHADER_STAGES_EN=0x482030: fast launch other than one point per subgroup is unsupported (State.cpp). The draw is an NGG fast-launch GS with VGT_PRIMITIVE_TYPE 6 (triangle strip), GE_CNTL 0x201 (one vertex, one primitive per subgroup). The only difference from main's supported case is the input primitive. We got past it with a local change that treats the input as a point list, which is a guess at the hardware's behaviour, so it isn't a PR.
2. sceAcmBatchJobPriority not implemented: feat(libSceAcm): implement sceAcmBatchJobPriority #7064.
3. layered draws into a color view of several array slices are unsupported: feat(agc): render layered draws into 2D array color views #7254. That run showed the character and the HUD, with the world a black rectangle.
4. VK_ERROR_DEVICE_LOST 78-255 s after New Game, in 8 of 9 runs; the ninth died by an access violation in nvoglv64.dll on the queue worker (descriptor-set allocation) instead. The first scene draws partly, at 0.4-1.9 FPS. GPU checkpoints put the last marker right before the dispatch of compute program 0x1004276f00. That shader also runs in the menus, so the fault depends on the scene's data. The device-fault report has no address infos, only "instruction pointer unknown" entries. That doesn't look like a page fault. A static read of that shader's IR found no unguarded pointer read. fix(agc): reuse a recycled descriptor set only for a table of its own size #6567, fix(agc): fault image table entries the device or guest memory cannot hold #6481 and fix(agc): prepare a draw or dispatch again when the memory it captured changed before it was recorded #7446 don't change it.

If your RADV runs get to New Game, it would be useful to know whether the device loss shows up on AMD too, or is NVIDIA-specific.

AI-assisted: yes (Claude Code)

[hugodearaujo]

Yes, it shows up on AMD too. Arch Linux, AMD Radeon RX 9070 XT, RADV (Mesa 26.2.5), main 48d61f5 with #7064, #7254 and a local change that treats the fast-launch input as a point list (the same guess as yours), first run with an empty shader cache.

After First Time Setup, the title screen and the main menu (about 25 FPS), New Game played its opening scene at about 6 FPS, and about 8 minutes after New Game the device was lost. On AMD it is a page fault. amdgpu logged [gfxhub] page fault and then ring gfx_0.0.0 timeout, and RADV printed:

radv: GPUVM fault detected at address 0x108004397000.
GCVM_L2_PROTECTION_FAULT_STATUS: 0x601431
         CLIENT_ID: (SQC (data)) 0xa
         PERMISSION_FAULTS: 3
         MAPPING_ERROR: 0
         RW: 0
[device-fault] read invalid at 0x108004397000 (precision 0x1000)

So a shader's scalar data read (SQC) hit a page it may not read. In the seconds before, AnyPS5 logged deferred write-back failed: image table: shader 0x1004138500 stage 2 pc 0x78 selected the descriptor at 0x3affcc580: outside the snapshot (four times), a BDA access failed: address=0x491d79ffc and a WAIT_REG_MEM_64 timeout on queue 0x3d. I haven't tied any of them to the fault, and this is one run.